Goodfire

AI · San Francisco, US

AI interpretability research lab: tools to understand, debug, and design models.

Founders

Eric Ho (United States · ♂) — A Yale CS grad who built and scaled AI recruiting startup RippleMatch, he founded Goodfire to make AI mechanistic interpretability practical and useful.
Dan Balsam (United States · ♂) — RippleMatch's founding engineer and Head of AI who deployed LLMs in production, he co-founded Goodfire as CTO to build tools that decode how AI models think.
Tom McGrath (United Kingdom · ♂) — He co-founded DeepMind's interpretability team, then left Google to make interpretability useful by joining Ho and Balsam as Goodfire's Chief Scientist.

Shares the most investors with (number = shared investors):

Date	Round	Amount
Feb 2026	Series B Goodfire closed a $150 million Series B led by B Capital at a $1.25 billion valuation for its AI model interpretability tools.	$150M
Aug 2024	Seed Goodfire, which builds mechanistic-interpretability tools for generative-AI models, raised a $7M seed round led by Lightspeed Venture Partners with Menlo Ventures, South Park Commons and others.	$7M

Tap an investor to see what else they have backed.

Data is community-curated and may be incomplete.